Do Somali Minnesotans commit crime at higher rates?

January 15, 2026: Replication data: HERE.

Spoiler: Somali Minnesotans do not commit crime at higher rates.

Calculating the crime rate for just Somalis compared to non-Somalis in Minnesota is challenging because when people get arrested or convicted, police and courts in Minnesota do not record national origin in public records. Here I consider a noisy and conservative workaround. The majority of Muslims in Minnesota are Somali and virtually all Somalis are Muslim or of Muslim heritage.  Muslim names are distinctive—names like Farah, Hirsi, and Warsame, and distinct Arabic spellings like Elmi and Abozok stand out in records. By hand-coding names on Minnesota's county jail rosters, prison populations, and state offender registries, we can estimate relative risk ratios for Muslims and non-Muslims.

The Census Bureau's 2020 census doesn't ask questions about religion, but there is an 2020 "religion census" by the Association of Religion Data Archives that does estimate number of religious adherents at the county level. Where the religion census is implausibly low (i.e. when the Somali estimate for the 2020 decennial census is higher than the Muslim estimate in the religion census), we can use the Somali share of the county population from the 2020 decennial as a proxy. Although this method has limitations and probably over-estimates Muslim crime incidence, it provides a conservative approximation. 

Starting at the state level, I find Muslims are less likely than non-Muslims to be reported on criminal offender registries. I hand-coded Minnesota’s sex offender list (465 names) and methamphetamine offender list (451 names). Minnesota does not have a public offender registry for driving under influence, but Ohio does have a habitual offender registry. The Buckeye state is also home to the second-largest Somali population in the United States. I hand-coded Ohio’s registry of 4,105 names. Below are the results.

We can divide the ‘risk’ a Muslim from the state’s Muslim population is registered as offender (# muslims on offender list / # muslims in state) by the risk a non-Muslim from the non-Muslim population is registered as offender (# non-muslim on offender list / # non-muslims in state). In epidemiology, this is called a risk ratio (RR). RRs lower than 1 mean Muslims are less likely to be on the offender list than non-Muslims. This is just a point estimate, a number, so what if it was a bad draw? In statistics, we can model the sampling uncertainty using Wald log-risk-ratio, fitting a 95% confidence interval on the risk ratio. There are more computationally intensive ways of doing this like bootstrapping the confidence interval, but the substantive result is the same, and Wald CIs are canon. Note that the 95% CI estimation strategy treats the population denominators as known and fixed and models variance as distributed binomial. Note that the confidence intervals are only symmetric on a logarithmic scale.

We find that Muslims are either statistically indistinguishable from their peers or less likely to be offenders:

We can do the same analysis for the jail rosters in Minnesota counties with significant Somali population, such as Stearns county with its seat St. Cloud (5% Somali per 2020 census), Anoka county with its steat Columbia Heights (8.8% Somali), Ramsey county, which has the capital city St. Paul (3.1% Somali) and Dakota county with Burnsville (5.9% Somali). Kandiyohi county, with its seat of Willmar city (9.5% Somali) had only 25 people in jail at the time I looked, but it does gives us warrant data for 933 people, so I use that.

The jail rosters for Hennepin and Ramsey counties, home of Minneapolis-St. Paul and the Cedar Riverside 'Little Mogidishu' neighborhood are more complicated because their police seem to arrest across county lines often, the jails receive arrests from the international airport and take transfers from other counties in preparation for state district court trials. Thus it’s unclear what the relevant population denominator should be for these carceral hubs. However, both Hennepin and Ramsey have county prisons for those county residents convicted and incarcerated. These data are actually better for our question because they contain people convicted of a crime, not just accused.

As you can see, magnitudes are similar. The table shows the percent of Muslims among the arrested and incarcerated is similar to their share in the population. The risk-ratios, taking into account sampling uncertainty, also confirm that Muslim risk of arrest and imprisonment is lower or statistically indistinguishable from that of non-Muslims in 5 out of 6 of the counties I looked at:

There are good reasons to think that the real risk ratio is even lower for Muslims than is reflected in this graph. First, many of the names coded Muslim are common Arab names, but most Arab-Americans (63%) are Christian. I did not include county-level Arab Christian estimates in the population denominator because they are unavailable. Inclusion would water-down the percentage further. Second, incarcerated people convert to Islam and adopt Muslim names at higher rates than the general population. For historic reasons, this is particularly true among black men. But converts are not the population of interest.  

Third, lest one dismiss these results as driven by a “soft on Somali crime” approach, police tend to have more coercive and surveillance resources in cities, where non-white minorities like Somalis live. Within cities, they tend to police Black neighborhoods, racial boundaries, and Black persons more often than predicted by race-specific estimates of crime participation. Minneapolis police specifically are no woke pansies; they inflict appalling racial violence. Recall for instance that a cop in 2020 murdered a Black man, George Floyd, in an act so brazen in its racism and cruelty, it provoked a national uprising that was only stopped with 14,000 arrests and the combined deployment of 96,000 soldiers. Both Minnesota’s Department of Human Rights and the U.S. Department of Justice released reports in 2023 from two-year probes into MPD describing “pattern or practice of race discrimination.” According to 2022 data, they were also the fourth most likely to use physical force in the entire United States. 

It is thus remarkable that these risk ratios, unadjusted for factors like age or poverty or more aggressive policing, still show that Muslims in Minnesota are no more likely than anyone else to wind up in jail or prison. Of course, it is possible that some Muslims on the list may evade my Muslim name detector (“type 2 error”)—it’s impossible to truly know. In any case, the burden of proof is on the right-wing, and they have not met it.